Newsletter 2021.1 Index
Theme : "Mechanical Engineering Congress, 2020 Japan (MECJ-20)”
In the development of Large-Eddy Simulation (LES), the monumental ideas have been proposed decennially in the previous century: the eddy viscosity model by Smagorinsky (1963), the theory of filtering by Leonard (1974), the scale-similarity model by Bardina, et al. (1980) and the dynamic eddy-viscosity model by Germano, et al. (1991). The success of LES for turbulent flow in a plane channel with the wall-model by Deardorff (1970) and the non-slip boundary condition by Schumann (1975) created expectations of new trends in the numerical investigation of turbulent flows. In fact, LES has become a practical tool for the numerical simulation of turbulence in industry(1). On the other hand, essentially new idea is not found in this century. However, it does not mean there is not unsolved issue in LES. To enforce the nonslip boundary condition is still expensive especially for high Reynolds number conditions. For this issue, the refinement to treat the near-wall turbulence has been extensively conducted and there seems favorable developments recently. More serious problem is involved in the basic equations. For example, the commutation of filtering and differentiation has been recognized from the beginning but it is still left neglected due to the mathematical difficulty. The imperfectness in the governing equation could become hidden barrier for the further development that including the application to multiphase flows that needs volume averages of multiple scales(2). Meanwhile, the introduction of machine learning might be unavoidable in the LES modeling especially for the purpose obtaining a prompt answer. The necessity of physics-informed machine learning approach has been already pointed out. In addition, the necessity of properly volume-filtered equations should be reaffirmed as the basis of novel development.
Large-Eddy Simulation, Volume Filtering, Commutation Error, Turbulent Flow, Computational Fluid Dynamics